| Claude Opus 5.5 | Phi 4 | |
|---|---|---|
| Input, per 1M tokens | $4.00 | $0.07 |
| Output, per 1M tokens | $20.00 | $0.14 |
| Context window | 1M | 16K |
| Classification or routing (600 tokens in, 10 out), per 1,000 | $2.60 | $0.0434 |
| Chat reply (2,000 in, 400 out), per 1,000 | $16.00 | $0.196 |
| Retrieval-augmented answer (8,000 in, 500 out), per 1,000 | $42.00 | $0.63 |
| Confidence signal | no | no |
| Tool calling | yes | no |
| BANKING77, correct of 400 | 354 | not measured |
Which one should carry your traffic?
If Phi 4 answers your requests as well as Claude Opus 5.5, moving that traffic saves 99% on chat-shaped requests. Averages cannot tell you whether it does for your requests. Phi 4 does not expose a confidence signal through OpenRouter, so Gimbix would test it as a full replacement rather than a first answer.
Claude Opus 5.5 profile · Phi 4 profile · Measure it on your traffic